Fermi reports $486 million loss in first year of operations

Fermi reported a net loss of $486.4M for its Jan 10–Dec 31, 2025 inception period and generated no revenue as Project Matador remains in development. The company completed an Oct 2025 IPO raising $745.6M (net) and held $408.5M in cash at Dec 31, 2025, while incurring $177.8M G&A (including $132.7M share‑based comp) and one‑time charges ( ~$173.8M charitable donation, $111.6M fair value adjustments). Project Matador targets up to 11 GW (expandable to 17 GW) on a 5,236‑acre site and received a 6 GW natural gas air permit in Feb 2026; the company has ~ $785M in equipment financings and a $156.3M senior note facility (undrawn). Fermi intends to elect REIT status starting with the short tax year ended Dec 31, 2025.

Analysis

Large-scale AI campus builds create two simultaneous markets: durable long-duration demand for compute and short-term capital and construction risk. OEMs that can pre-sell or flex production of servers and racks (SMCI-style supply chains) will capture the early revenue tail, while developers that carry construction and permitting risk will see highly variable returns until take-or-pay leases crystallize.

Key catalysts are not market sentiment but binary execution milestones — signed long-term leases, transmission interconnect capacity agreements, and final environmental approvals. These events work on different timelines (weeks-months for leases, quarters for interconnect studies, years for expanded permitting) and each can swing valuation multiples far more than near-term macro noise.

Actionable positioning should separate exposure to pure compute demand from developer/execution risk. Use asymmetric option structures to express views: buy convexity on server OEMs and hedge or limit downside on the developer through puts or short exposure sized to execution risk. Capital structure arbitrage (junior equity vs secured equipment finance) is also fertile: non-dilutive financing lines reduce immediate equity risk but raise refinancing/ covenant risk at future maturities.

Contrarian angle: the market’s “buy the dip” instinct for a marquee AI campus underprices two offsetting effects — upside optionality from land and REIT-style cash yields once stabilized, and downside from multi-year buildout, permit creep and refinancing windows. Investing around milestone triggers (lease announcements, interconnect clearances) offers cleaner payoff profiles than owning the equity through the full build cycle.

Key Decisions for Investors

• Long SMCI via 6–12 month call spread (buy calls, sell higher strike) sized to 1–2% portfolio — thesis: capture server/order re-acceleration if large tenants commit; downside = option premium, upside ~2x–4x for a moderate cost.
• Hedge FRMI execution risk with a 9–15 month put-spread (buy puts, sell lower strike) sized to offset 25–40% of equity exposure — protects against lease/permitting misses while keeping limited cost; payoff rises sharply on dilution or re-pricing events.
• Pair trade: Long SMCI / Short FRMI equal dollar for 6–18 months — expresses preference for compute demand capture vs developer execution; target 20–30% IRR if SMCI outperforms by 15–25%, stop-loss at 12% adverse move.
• Event-driven small long FRMI position (up to 1% portfolio) only after confirmed long-term tenant LOI or definitive lease — risk/reward improves post-signing as valuation re-rates toward stabilized REIT multiple; until then avoid large outright longs.